Download Mapping the distribution of dholes

Mammalia xx (2012): xxx–xxx © 2012 by Walter de Gruyter • Berlin • Boston. DOI 10.1515/mammalia-2011-0063
Mapping the distribution of dholes, Cuon alpinus
(Canidae, Carnivora), in Thailand
Kate E. Jenks1,2,*, Shumpei Kitamura3, Antony J.
Lynam4, Dusit Ngoprasert5, Wanlop Chutipong5,
Robert Steinmetz6, Ronglarp Sukmasuang7,
Lon I. Grassman Jr.8, Passanan Cutter9,
Naruemon Tantipisanuh5, Naris Bhumpakphan7,
George A. Gale5, David H. Reed10, Peter Leimgruber2
and Nucharin Songsasen2
1
Graduate Programs in Organismic and Evolutionary
Biology, and Wildlife and Fisheries Conservation,
University of Massachusetts, 611 North Pleasant Street,
Amherst, MA 01003, USA, e-mail: [email protected]
2
Smithsonian Conservation Biology Institute, National
Zoological Park, 1500 Remount Road, Front Royal, VA
22630, USA
3
The Museum of Nature and Human Activities, Hyogo,
Yayoigaoka 6, Sanda 669-1546, Japan
4
Wildlife Conservation Society, Global Conservation
Program, 2300 Southern Boulevard, Bronx, NY 10540,
USA
5
Conservation Ecology Program, School of Bioresources
and Technology, King Mongkut’s University of
Technology Thonburi, Bangkhuntien, Bangkok 10140,
Thailand
6
World Wildlife Fund for Nature-Thailand, Ladyao,
Jatujak, Bangkok 10900, Thailand
7
Department of Forest Biology, Faculty of Forestry,
Kasetsart University, Bangkok 10900, Thailand
8
Feline Research Center, Caesar Kleberg Wildlife
Research Institute, Texas A&M University-Kingsville,
MSC 218, 700 University Boulevard, Kingsville, TX
78363, USA
9
Conservation Biology Program, University of Minnesota,
187 McNeal Hall, 1985 Buford Avenue, St. Paul, MN
55108, USA
10
Department of Biology, University of Louisville,
Louisville, KY 40292, USA
*Corresponding author
Abstract
No recent attempt has been made to survey dhole distribution,
or to estimate remaining population numbers. We surveyed
15 protected areas in Thailand with camera traps from 1996
to 2010. We used the photo locations of dholes (n = 96) in the
maximum entropy (MaxEnt) model along with six environmental variables to model current dhole distribution, as well as
species predictive occurrence layers for sambar, red muntjac,
wild boar, tiger, and leopard. The MaxEnt model identified
the predicted probability of the presence of leopards and sambar as positive and the most important variables in modeling
dhole presence, indicating that maintaining a sufficient prey
base may be the most important factor determining continued survival of dholes. Roughly 7% of the total land area in
Thailand is potentially suitable for dholes. However, surveys
to date have focused on protected areas, which make up just
a third of the potential suitable areas for dholes. Only in four
protected areas do they occur across the entire landscape, suggesting that in the majority of places where they occur, habitats are not uniformly suitable. Using the model, we identified
several potential areas where dholes have not been reported,
and therefore status surveys are needed, and where future
research of the species might be focused.
Keywords: Cuon alpinus; MaxEnt; maximum entropy
modeling; Southeast Asia; species distribution modeling.
Introduction
The dhole Cuon alpinus (Pallas 1811) is a medium-sized
social canid that once occurred over a wide geographic
range from the Tian-Shan and Altai Mountains in central
Asia and easternmost Siberia to India and Indochina (Cohen
1978, Durbin et al. 2004). Although listed as endangered
[International Union for Conservation of Nature and Natural
Resources (IUCN) 2008], little is known about current dhole
population sizes and distribution across its current geographic
range. India is thought to be the current stronghold for dholes,
although existing information on the species primarily stems
from Johnsingh’s 1976–1978 field study in Bandipur and his
1985 census of dholes using questionnaire surveys (Johnsingh
1981, 1985), which only covered as far east as Myanmar and
is now out of date. In an attempt to map the dhole’s range
in 1993, Stewart (1993) conducted interview surveys across
Southeast Asia but was unable to locate dholes in Thailand.
The most up-to-date distribution map of this species was
compiled from status reports for the 2004 Canid Action Plan;
however, this map contains huge tracts of localities with
unconfirmed or unknown dhole status (Durbin et al. 2004).
No recent attempt has been made to survey dhole distribution
or to estimate remaining population numbers.
Our study intends to use established distribution modeling tools to develop and test the first ever distribution map
for dhole, an endangered canid species, in Thailand. The
results from our research are meant to assist conservation
decision makers for prioritizing geographic areas for future
dhole surveys, research, management, and conservation. To
achieve this, we developed a new and unique approach that
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
2
K.E. Jenks et al.: Thailand dholes
incorporates distribution models of prey and competitor species to significantly increase the predictive power of our distribution models (Anderson et al. 2002, Singh et al. 2009). As
Thailand has one of the most extensive protected area systems
in the region (Pattanavibool and Dearden 2002), understanding the dholes’ distribution in this country should have significant conservation benefits. Our modeling efforts will also
advance current understanding of the ecology of dhole and
the factors that control this species’ geographic distribution
in Thailand. Finally, we believe that our approach could be
translated to better explore the status and distribution of dhole
throughout its geographic range.
Several factors may influence patterns of dhole distribution, including vegetation and landscape structure, food
availability, competition with other carnivores, and human
population levels (e.g., Morris 1925, Prater 1965, Barnett
et al. 1980, Johnsingh 1983). However, a lack of understanding of specific influences on dholes of different environmental
and human factors is an impediment to managing their populations in Asian protected areas.
Vegetation type does not seem to constrain dhole distribution as they occur across a wide range of land cover types,
including tropical dry and moist deciduous forest, evergreen
and semi-evergreen forests, low scrub interspersed with bamboo, grasslands, and alpine steppe (e.g., Peploe 1947, Barnett
et al. 1980, Johnsingh 1981, 1983); dholes even inhabit open
country in Ladak and Tibet (Prater 1965). Johnsingh (1981)
suggested that dholes may prefer open forest to dense forest
and the moist deciduous forests of India may represent optimal habitats (Phythian-Adams 1949). In an Indian study that
applied occupancy models to data from country-wide experts,
the best model for predicting dhole occupancy included covariates for evergreen, temperate, and deciduous land cover
with low and mid elevations (Karanth et al. 2009). However,
it is possible that dholes’ apparent vegetation preferences are
actually the result of prey distributions and avoidance of competing predators (Johnsingh 1981).
Dhole prey selection varies throughout the range, but they
often tend to focus on medium to large ungulates. Sambar
(Kerr 1792) (Rusa unicolor), Wild boar (Linnaeus 1758)
(Sus scrofa), Tahrs (C.H. Smith 1826) (Hemitragus jemlahicus), Muntjac (Muntiacus spp.), Chital (Erxleben 1777)
(Axis axis), Markhors (Wagner 1839) (Capra falconeri),
Musk deer (Moschus spp.), and Goral (Naemorhedus spp.)
have all been recorded among dhole prey items (e.g., Morris
1925, Prater 1965, Barnett et al. 1980, Johnsingh 1983). Yet,
in Mudumalai Wildlife Sanctuary in India, hares and rodents
comprised 46% of the dholes’ diet (Barnett et al. 1980), while
in Taman Negara National Park in Malaysia, 78% of dhole
scats contained mouse deer [(F. Cuvier 1822) (Tragulus napu)
and (Osbeck 1765) (Tragulus javanicus); Kawanishi and
Sunquist 2008]. This indicates that dholes may be able to rely
on smaller prey items in areas where ungulate populations
have declined.
Dholes are sympatric with tigers (Linnaeus 1758) (Panthera
tigris), leopards (Linnaeus 1758) (Panthera pardus), and
jackals (Linnaeus 1758) (Canis aureus) throughout Southeast
Asia (Johnsingh 1992) and with wolves (Linnaeus 1758)
(Canis lupus) in China and India (Johnsingh and Yoganand
1999), begging the question of whether dholes compete with
other carnivores for shared prey. Johnsingh (1992) identified
13 parameters that enabled tiger and dholes to coexist, and
partitioning of prey selection was identified as the top factor
(Karanth and Sunquist 2000). Although prey partitioning may
enable coexistence of tigers and dholes, interguild predation,
i.e., direct predation of the smaller by the larger carnivore,
may lead to greater separation. In our case, this would mean
the distribution and abundance of larger, potentially competing carnivore species may restrict dhole habitat selection
(Woodroffe and Ginsberg 2005). Examples of this include
cases where dholes have been killed by tigers and attacked by
leopards (e.g., Johnsingh 1983, Karanth and Sunquist 2000,
Lynam et al. 2001), indicating both larger carnivores may be
behaviorally dominant over dholes. Moreover, Venkataraman
(1995) argues that dholes need to be aggressive toward
leopards as a defense against leopard attacks on dholes.
Other dhole habitat considerations include suitable denning
sites and proximity to water (Inverarity 1901, Prater 1965),
although no study has suggested that den sites are a limiting
resource. Dholes are known to hunt sambar by driving them
into water bodies (Johnsingh 1983). Dhole distribution may
be inversely related to human distribution because the species
is sometimes persecuted as livestock predators, prey populations are often reduced by humans, and domestic dogs may
transmit diseases (Durbin et al. 2004).
Dholes overlap with other large carnivores throughout
their range; however, they probably play a unique ecological role that is not functionally redundant with the roles
of other carnivores (Woodroffe and Ginsberg 2005). This
implies that dholes have their own unique impacts on prey
species and ecosystem processes, and that their conservation is important for maintaining ecological function and
community integrity. To explore this influence and better
elucidate the ecological role of dholes, managers first need
to understand where the species occurs and why. We used
data collected from 1996 to 2010 from 15 protected areas to
assess potential factors affecting the distribution of dholes
in Thailand. Our goals were to (1) confirm the presence of
dholes in protected areas in Thailand, (2) identify environmental factors associated with dhole occupancy, (3) predict which areas within the country are within the species’
potential distribution, (4) evaluate the efficacy of protected
areas in Thailand in providing sufficient area for viable
dhole populations, and (5) identify areas for future research
efforts on this endangered species.
Materials and methods
Input data
From 1996 to 2010, camera traps were deployed at 15 protected areas within Thailand: Bang Lang National Park (BL),
Hala-Bala Wildlife Sanctuary (HB), Huai Kha Kaeng Wildlife
Sanctuary (HKK), Kaeng Krachan National Park (KK), Khao
Ang Rue Nai Wildlife Sanctuary (KARN), Khao Sam Roi
Yod National Park (KSRY), Khao Sok National Park (KOS),
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
K.E. Jenks et al.: Thailand dholes
N
PK
KY
THP
HKK
TYW
MP
TAP
KK
KARN
KB
KSRY
KLS
Camera trap survey
Protected area
KOS
0
BL
125
250
500 km
HB
Figure 1 The study locations included 15 protected areas within
Thailand: Bang Lang National Park (BL), Hala-Bala Wildlife
Sanctuary (HB), Huai Kha Kaeng Wildlife Sanctuary (HKK), Kaeng
Krachan National Park (KK), Khao Ang Rue Nai Wildlife Sanctuary
(KARN), Khao Sam Roi Yod National Park (KSRY), Khao Sok
National Park (KOS), Khao Yai National Park (KY), Klongsaeng
Wildlife Sanctuary (KLS), Kuiburi National Park (KB), Maenam
Pachi Wildlife Sanctuary (MP), Phu Khieo Wildlife Sanctuary (PK),
Ta Phraya National Park (TAP), Thap Lan National Park (THP), and
Thung Yai Naresuan-West Wildlife Sanctuary (TYW). Dholes were
detected at underlined sites.
Khao Yai National Park (KY), Klongsaeng Wildlife Sanctuary
(KLS), Kuiburi National Park (KB), Maenam Pachi Wildlife
Sanctuary (MP), Phu Khieo Wildlife Sanctuary (PK), Ta
Phraya National Park (TAP), Thap Lan National Park (THP),
and ThungYai Naresuan Wildlife Sanctuary-West (TYW;
Figure 1). In total, individual cameras were set at 1174 sites,
and accumulated 48,130 trap nights with a mean of 41 trap
nights per camera. Camera trap sites were not baited and
placed a minimum of 0.5 km apart at elevations ranging from
0 to 1351 m (mean 428 m). All cameras were operational
24 h per day, and recorded time and date for each exposure.
Cameras were placed ∼50 cm from the ground and close to
trails, stream beds, and ridges where wildlife signs (i.e., foot
prints and scats) were present to maximize the chances of
capturing an animal.
Habitat variables
We used six environmental variables across Thailand, together
with the predicted occurrence of three prey species, and the
3
predicted occurrence of potential competitors (tigers and leopards), to model dhole distribution. For assessing the potential
distribution of prey and competitor species, independent probability of occurrence models were developed using maximum
entropy (MaxEnt). Following the modeling procedure for
dholes described below, we first produced predicted occurrence layers for three prey species (Sambar, Red muntjac, and
Wild boar) on the basis of locations where the species were
photo-trapped. The predicted occurrence layers are a surrogate
for prey availability or abundance. The resulting layers were
included as predictor variables in the tiger, leopard, and dhole
models. Additionally, the output layer for Tiger and Leopard
was included as a variable for the dhole model. Presence
records used for training included sambar (n= 124), red muntjac (n = 271), wild boar (n= 184), tiger (n = 80), and leopard
(n = 100). We are assuming that the inclusion of prey and competitors linked with environmental variables (Anderson et al.
2002, Singh et al. 2009) will result in a better spatial representation of the distribution of dholes, which cannot be substituted
by only using environmental variables. Because MaxEnt does
not use a statistical regression, but an optimization, colinearity and correlation of variables are not used in the theoretical
analysis of MaxEnt. It is therefore more stable regarding correlated variables (Elith et al. 2011) and unnecessary to create
separate abiotic and biotic models.
Annual precipitation (1950–2000) and elevation was
obtained from the WorldClim database (version 1.4, http://
www.worldclim.org). General country-wide land cover categories and distances to nearest protected area edge, village,
and stream were obtained using ArcMap 9.3 from shape files
provided by the Thailand Department of National Parks,
Wildlife, and Plant Conservation. Distance to nearest edge
was measured from all grid points within a protected area
to the boundary, with areas outside of protected areas being
assigned a zero value. We consolidated land cover categories
from 25 to 14; they were entered as a categorical variable
in the model and included agriculture, bamboo, beach forest, dry dipterocarp forest, dry evergreen forest, eucalyptus
plantation, grassland, hill evergreen, mixed deciduous forest,
moist evergreen forest, pine forest, secondary growth forest,
teak plantation, and other.
Distribution modeling
Locations at which dholes were photo-trapped (n = 96) were
the source of data for MaxEnt (Phillips et al. 2006). MaxEnt
estimates a frequency distribution by finding the distribution
that is closest to uniform, constrained by the average values for a set of variables taken from the target distribution
(Phillips et al. 2006). We used MaxEnt because it performs
better than other presence-only modeling techniques (Elith
et al. 2006), especially with low numbers of occurrence locations (Papes and Gaubert 2007). This method has been used
to develop habitat suitability models for a range of mammals
(e.g., DeMatteo and Loiselle 2008, Monterroso et al. 2009,
Wilting et al. 2010, Jennings and Veron 2011).
All environmental layers were projected to the Indian
1975 UTM zone 47N to match their coordinates, clipped
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
4
K.E. Jenks et al.: Thailand dholes
to the extent of the boundary of Thailand, resampled to the
same cell size of 30 arc-seconds (∼1 km2), and entered with
the occurrence data into MaxEnt version 3.3.3 (http://www.
cs.princeton.edu/~schapire/MaxEnt). We set the program to
run 500 iterations with a convergence threshold of 0.00001,
a regularization multiplier of 1, a maximum of 10,000 background points, the output grid format as “logistic,” algorithm
parameters set to “auto features,” and all other parameters at
their default settings (Phillips and Dudik 2008). The model
was trained usinga mask to include surveyed protected
areas only because we only sampled camera-trap locations.
The final distribution map resulted from the model projecting into all of Thailand, including protected areas where we
collected no data and outside of protected areas. The model
was trained using a mask to include surveyed protected areas
only because we only sampled camera-trap locations for prey,
competitors, and dholes within protected areas. In this way,
our background sample excluded areas that have not been
searched (Elith et al. 2011). While extrapolation beyond the
area where the data was collected often is an issue in distribution modeling, MaxEnt has consistently preformed very
well in such applications. We experimented with different
threshold values; however, this tends to lead to significant
overprediction. On the basis of our experience and published
literature, we feel justified in our approach. The outputs also
are parsimonious and can serve as a first hypothesis for areas
where we may find additional dhole populations or may want
to consider potential sites for dhole restoration and recovery.
We had the program randomly withhold 25% of the presence
locations to test the performance of each model.
Model performance was assessed by the area under the
curve (AUC) of the receiver-operating characteristic (ROC)
plot (Liu et al. 2005). We calculated standard errors and confidence intervals for each of the models using ROCR (Sing et
al. 2009), vcd (Meyer et al. 2010), and boot (Canty and Ripley
2010) packages in R v2.11.1 (R Development Core Team
2010). The data were jackknifed to evaluate each variable’s
importance in explaining the observed distribution. The percent contribution of each variable was calculated on the basis
of how much the variable contributed to an increase in the
regularized model gain as averaged over each model run. To
calculate variable permutation, for each variable in turn, the
values of that variable on training presence and background
data were randomly varied and the resulting change in training
AUC is shown normalized to percentages (MaxEnt Tutorial;
http://www.cs.princeton.edu/∼schapire/MaxEnt/). To delineate
areas with better than random prediction of dhole presence, we
used a priori prevalence values (96 dhole detection locations
out of 1174 total sampling locations) as the “presence” threshold (0.08 detection). In a review of 12 approaches for choosing
a threshold of occurrence, Liu et al. (2005) ranked this prevalence approach as one of the most robust.
Results
Distribution models for sambar, red muntjac, wild boar, tiger,
leopard, and dhole performed well based on the moderately
high ( > 0.80) AUC values (Swets 1988; Table 1). Distance to
protected area edge (Edge) and annual rainfall (Rain) had the
highest predictive power for all prey species (Table 2). The
probability of presence for prey was higher at lower rainfall
locations and at distances closer to the interior of protected
areas (graphs not shown).
The variables with the highest percent contribution and
permutation importance for the dhole model were Leopard
and Sambar (Table 2). The jackknife test of variable importance shows the highest gain when the variable Sambar is
used in isolation, which therefore appears to have the most
useful information by itself (Figure 2). However, the variable
that decreases the gain the most when it is omitted is Leopard,
which indicates this variable has the most information that is
not present in the other variables (Figure 2). The variables
Leopard, Sambar, Stream, and Red muntjac make up almost
90% of the contribution for the dhole model, and all of the
variables are positively correlated with predicted dhole presence (Table 2, Figure 3).
We generated a map (predicted probability of occurrence;
Figure 4) of potential dhole distribution in Thailand. The highest dhole probability of presence was projected to be < 500 m
elevation (graph not shown). The land cover categories with
the highest predicted probability of dholes as calculated
in MaxEnt included grasslands (predicted dhole presence
of 60%), mixed deciduous forest (57%), dry dipterocarp
forest (49%), dry evergreen forest (47%), and hill evergreen
forest (29%; Figure 5). The model predicted < 10% probability of occurrence of dholes in all other land cover types.
The total area predicted to be potential habitat for dholes was
34,404 km2, which is roughly 7% of the total area in Thailand
(Figure 6).
Thirty percent of this potential habitat for dholes falls
within current protected areas. If we exclude all land
outside of protected areas, the total potential habitat for
dholes is 10,461 km2 or 2% of Thailand (Figure 6). To
further refine the potential habitat, we excluded patches
that are too small to support a dhole pack. Grassman et al.
(2005) found dholes in PK to have ranges of 12.0 and 49.5
km2. Therefore, we counted only contiguous patches of
predicted habitat > 50 km2. The remaining 31 patches have
the potential to support 161 dhole home ranges of 50 km2 on
the basis of our rough assumptions. However, 58% of those
patches might sustain fewer than three packs (Figure 7).
From the model, we identified four protected areas [Khlong
Table 1 Model of predicted species occurrence output from
MaxEnt; performance measured by the AUC of the ROC plot,
standard error (S.E.), and 95% confidence interval (95% CI).
MaxEnt model
AUC
S.E.
95% CI
Sambar
Red muntjac
Wild boar
Tiger
Leopard
Dhole
0.883
0.827
0.827
0.715
0.929
0.932
0.025
0.022
0.026
0.060
0.040
0.033
0.827 – 0.927
0.782 – 0.880
0.772 – 0.874
0.589 – 0.843
0.827 – 0.981
0.850 – 0.993
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
K.E. Jenks et al.: Thailand dholes
5
Table 2 Estimates of relative percent contribution (RC) and permutation importance normalized to percentages (PI) for variables used in
MaxEnt modeling of species distributions in Thailand.
Variables
MaxEnt model
Sambar
RC
Leopard
Sambar
Stream
Red muntjac
Elev
Land cover
Village
Wild boar
Tiger
Edge
Rain
PI
Red muntjac
Wild boar
RC
RC
PI
Tiger
PI
2.1
0.7
2.4
3.2
10.2
4.9
10.7
14.1
10.7
10.8
9.1
24.9
11.2
9.8
6.6
6.1
9.7
13.7
10.1
14.8
4.1
9.7
8.4
12.8
35.1
27.4
26.4
28.1
29.2
40.7
33.6
33.7
19.2
41.6
33.7
30.6
Leopard
Dhole
RC
PI
RC
PI
35.0
5.2
7.5
2.3
16.4
13.9
16.7
2.9
0.5
2.1
1.8
31.9
18.1
16.3
29.7
0.7
7.0
2.5
26.9
5.0
10.8
3.7
0
0.4
3.5
22.1
4.1
27.0
0.5
2.6
1.5
25.0
1.5
15.8
1.0
38.2
RC
PI
37.6
36.8
8.5
6.9
4.0
2.5
1.3
1.2
0.7
0.4
0.2
39.2
25.3
5.0
10.5
3.8
8.3
2.4
0
0.5
4.1
0.9
MaxEnt, maximum entropy; Edge, distance of species presence to protected area boundary (m); Elev, elevation (m); Land cover, categorical
land cover type; Rain, annual rainfall (mm); Stream, distance of species presence to nearest stream (m); Village, distance of species presence
to nearest village (m); Sambar, Red muntjac, Wild boar, Leopard, Tiger, predicted layer of occurrence from MaxEnt modeling for each species.
Variables are in italic.
Wangchaow National Park (KW), Salakpra Wildlife
Sanctuary (SP), Khao Ang Rue Nai Wildlife Sanctuary
(KARN), and Pang Sida National Park (PS)] where dholes
may range across almost the entire protected area and that
include patches > 50 km2 (Figure 8).
Rain
Edge
Tiger
Predictor variable
Wild boar
Village
Land cover
Elev
Red muntjac
Stream
Sambar
Leopard
Total
0
0.5
1.5
1
2
Regularized training gain
Total
With only variable
Without variable
Figure 2 Jackknife analyses of individual predictor variables
important in the development of the full model for dholes in relation to the overall model quality or the “regularized training gain.”
Black bars indicate the gain achieved when including only that variable and excluding the remaining variables; gray bars show how
much the gain is diminished without the given predictor variable.
Edge, distance to protected area boundary (m); Elev, elevation (m);
Land cover, categorical land cover type; Rain, annual rainfall (mm);
Stream, distance to nearest stream (m); Village, distance to nearest
village (m); Sambar, Red muntjac, Wild boar, Leopard, Tiger, predicted layer of occurrence from MaxEnt modeling for each species.
Discussion
Identifying areas of high habitat suitability for dholes lays
the foundation for planning future research and conservation
initiatives. Our MaxEnt results are a step toward highlighting areas of suitable habitat for dholes. We extrapolated our
predictions to the whole of Thailand to identify areas that
may be suitable for dholes but were previously disregarded.
We incorporated competitors because interactions may skew
dhole distribution despite environmental suitability and prey
availability. While we recognize the potential circularity of
the model because we used the same environmental variables
to develop distribution models that were surrogates for prey
availability and competitor presence, we elected to proceed
because so little is known about dholes and we thought it
was important to include these covariates, limited as the data
may be. Not surprisingly, prey availability (Sambar and Red
muntjac combined) explained 44% of the species’ predicted
occurrence.
The probability of presence for prey was higher at lower
rainfall locations and at distances closer to the interior of protected areas. These findings are dissimilar to Ngoprasert et al.
(in press) who found sambar and wild boar associated with
higher rainfall. However, they found red muntjac associated
with areas far from forest edges.
Our results indicate that the strongest correlate with the
distribution of dholes, which led to the highest model gain
when used in isolation, is the presence of Sambar. A strong
association with this single prey species was expected considering sambar comprise 30% of the frequency of occurrence of
prey items in dhole feces in Thailand (Grassman et al. 2005,
Salangsingha and DoungKae 2009) and up to 90% occurrence
in feces of dhole in India (Rice 1986).
Besides prey, the other variable shaping dhole distribution is the presence of leopards. Although there are accounts
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
K.E. Jenks et al.: Thailand dholes
A
1.0
N
0.8
0.6
0.4
0.2
0.0
0.0
Dhole probability of presence
B
Dhole probability of presence
0.4
0.8
0.6
1.0
Leopard probability of presence
1.0
0.8
0.6
0.4
0.2
Probability
0.0
0.0
C
0.2
High
0.2
0.4
0.6
0.8
400 km
Low
0.8
0.6
Figure 4 Predicted distribution for dholes within Thailand estimated
by MaxEnt modeling. Potential areas are shown in gray shading, with
the darker color indicating higher probabilities of occurrence.
0.4
0.2
0
5000
10,000
15,000
20,000
Distance to nearest stream (m)
Dhole probability of presence
100 200
Sambar probability of presence
1.0
0.0
D
0
1.0
1.0
0.8
0.6
two species share habitats owing to similar prey preference
(e.g., Johnsingh and Yoganand 1999, Karanth and Sunquist
2000). Sambar (a large prey, on average > 180 kg) contributed most in the MaxEnt model for leopards and second most
for dholes.
The other possibility is that the positive association of
dholes with leopards is related to their predicted negative
association with tigers, their potential intra-guild predator.
0.4
1.0
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Red muntjac probability of presence
Figure 3 Graphical representation of the relationship between
Leopard (A), Sambar (B), Stream (C), Red Muntjac (D), and Dhole
probability of presence. Each of the curves represents a different
MaxEnt model created using only the corresponding variable.
of interspecific competition between leopards and dholes
(Johnsingh 1983, Wood 1929, Venkataraman 1995), our
modeling indicates that dhole presence increased as leopard presence increased. This probably arises because these
Dhole probability of presence
Dhole probability of presence
6
0.8
Grassland
0.6
Mixed deciduous
Dry dipterocarp
0.4
Dry evergreen
Hill evergreen
0.2
0.0
Land cover type
Figure 5 Land cover categories with the highest probability of predicted dhole presence in Thailand.
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
K.E. Jenks et al.: Thailand dholes
N
Dhole absent
Dhole present (inside protected area)
Dhole present (outside protected area)
0
80 160
320 km
Figure 6 Predicted occurrence of dholes in Thailand on the basis
of a 0.08 threshold of prevalence (96 dhole detection locations out
of 1174 total sampling locations). Areas in black represent predicted
dhole occurrence inside protected areas. Areas in gray represent predicted dhole occurrence outside of protected areas.
However, our Tiger variable as modeled contributed very
little (neither positively nor negatively) to predicted dhole
distribution.
The potential range of dholes covered a wide spectrum
of habitats; however, our model predicted that dholes occur
primarily in grasslands and mixed deciduous forest at generally 150 m elevation. This is consistent with a previous radiotelemetry study in PK where one dhole pack was found to
10
Number of patches
9
8
7
6
5
4
3
2
1
0
1
2
3
5
6
7
8
10
12
14
32
Number of dhole packs in patch
Figure 7 Approximate number of dhole packs in a given patch vs.
the number of patches.
7
base its home range around a grassland area (Grassman et al.
2005). We caution, however, that this may be an artifact of
the prey distribution in PK and may not represent the general
population. Regardless, overall, the impact of land cover type
alone only contributed 2.5% to predicting dhole occurrence;
this emphasizes that prey base, not land cover type, is the
main limiting factor for the species.
The range of predicted dhole habitat does not expand into
the north and the model failed to predict dhole occurrence
in Doi Chiang Wildlife Sanctuary and Lum-nam-pai Wildlife
Sanctuary in the far north, despite field records of dhole sign
from Kanchanasaka (2005). The protected areas in this region
may preclude large populations of sambar due to poaching
pressure (Pattanavibool and Dearden 2002). However, the
north does have contiguous forest cover; we could be missing key variables or there is the possibility that poor input
(e.g., an out-of-date land cover layer) may have resulted in the
model being inaccurate for this region.
This study provides a first indication of how much dhole
habitat is not protected in Thailand. Our results show that
currently only 30% of potential habitat falls within protected
areas in Thailand. Additionally, protection measures inside
protected areas may not be adequate because there are large
areas of potential habitat inside protected areas where dholes
are apparently absent. This might be related partly to prey
availability, but the edge itself may be a sink for dholes due
to the proximity to human settlements and the greater likelihood of getting shot (K. Jenks, pers. obs.) or poisoned
(S. Vitnitpornsawan, pers. obs.) there relative to the safer core
area. The observation that wildlife abundance is higher in
central parts of protected areas vs. marginal areas has been
specifically documented in KY (Lynam et al. 2003, Jenks
et al. 2011). Current protection efforts are most intense in
areas close to a park or sanctuary headquarters, with remote
areas getting less protection. We recommend that protected
area edges be specially managed to support dhole and their
prey.
Additionally, there is a low probability that the 70% of
potential dhole habitat outside of protected areas actually supports dholes because there are no verifiable records of dholes
living outside of protected areas in Thailand. Many areas predicted to have suitable ecological conditions for dholes may
actually be devoid of dhole populations because virtually
all forests outside of protected areas in Thailand have been
converted to agriculture and intersected by roads for human
settlement. If forests are still present, they are likely to be
largely without prey. For example, even sambar is now listed
as vulnerable due to intense poaching pressure (IUCN 2008).
Moreover, the lack of formalized protection measures outside of protected areas means that dhole survival chances are
much reduced there.
If we exclude all areas outside of protected areas, the
total potential habitat for dholes is 10,461 km2 or only 2%
of Thailand. If we restrict these areas further to include only
contiguous patches of predicted habitat > 50 km2, the remaining patches very roughly support 161 dhole home ranges.
Another challenging issue facing individuals involved in
dhole conservation is locating suitable sites for basic research
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
8
K.E. Jenks et al.: Thailand dholes
N
KW
PS
KARN
SP
Protected area
Predicted absence
0 20 40 km
Predicted presence
0
45
90
180 km
Figure 8 Individual protected areas in Thailand that are recommended for future field studies based on the large area of dhole presence:
Khlong Wangchaow National Park (KW), Salakpra Wildlife Sanctuary (SP), Pang Sida National Park (PS), and Khao Ang Rue Nai Wildlife
Sanctuary (KARN). Predicted presence of dholes on the basis of a 0.08 threshold and only includes patches that are > 50 km2.
of this elusive species. Our MaxEnt predictive map is a preliminary step that can guide field research to further clarify
the breadth of the dhole’s distribution. The model can be
tested by surveying for dhole presence in (1) areas predicted
to have a high probability of dholes and (2) areas predicted to
have no probability of dholes where the model may be wrong.
We identified four protected areas (KW, SP, PS, and KARN)
where dholes were predicted to range across almost the entire
area. These are ideal starting locations for basic dhole ecological research. However, these areas also represent locations where large predators (i.e., potential competitors) are
almost absent, especially tiger. To understand more about the
fate of dholes in the presence of large predators, research also
needs to be conducted in protected areas where dholes were
predicted to be present along with tigers and leopards, such as
KK, KB, HKK, and TYW. Additionally, we need to determine
if dholes are using areas outside of protected areas. The focus
for this should stem from areas of predicted presence from
our modeling, including the area north of MP and south of
SP, and a region southwest of PK. Finally, we also support
surveys in the north to test whether the model predictions are
correct in this region.
We have now explored the question of where dholes are
found; the next step is to shed light on the size and stabilities
of their populations. It may be that 7% of the country is potential habitat for dholes; however, this was inferred from a small
number of records and may not support stable populations,
but only small, isolated packs. Furthermore, we estimated that
the majority of contiguous patches may support fewer than
three packs. MaxEnt modeling has provided us with a helpful
evaluation of the distribution of the dhole in Thailand, which
can now be used for conservation planning.
Acknowledgments
This paper is dedicated to the memory of David H. Reed. We
wish to thank the Clouded Leopard Project, TRF/BIOTEC Special
Program for Biodiversity Research and Training Thailand, Kasetsart
University, and the Smithsonian Conservation Biology Institute for
providing funding for the initial conference that brought all the authors together to discuss collaboration. We are grateful to the following institutions for allowing us to use data collected under their
auspices: the Wildlife Conservation Society Thailand, World Wide
Fund for Nature – Thailand, and Smithsonian Institution. K. Jenks
was supported by a National Science Foundation Graduate Research
Fellowship, Association of Zoos and Aquariums Conservation
Endowment Fund, and a Smithsonian Endowment Fund. S.
Kitamura was funded by the Mahidol University Government
Research Grant, the National Center for Genetic Engineering and
Biotechnology, the Hornbill Research Foundation, and a JSPS
Research Fellowship. We thank J.W. Duckworth and A. Wilting for
their kind assistance in identifying the carnivores in our data set.
T.K. Fuller and two anonymous reviewers provided helpful comments on earlier drafts of the manuscript.
References
Anderson, R.P., A.T. Peterson and M. Gomez-Laverde. 2002. Using
niche-based GIS modeling to test geographic predictions of competitive exclusion and competitive release in South American
pocket mice. Oikos 98: 3–16.
Barnett, B.D., J.A. Cohen, A.J.T. Johnsingh and M.W. Fox. 1980.
Food habits of the Indian wild dog (Cuon alpinus): a preliminary
analysis. J. Bombay Nat. Hist. Soc. 77: 314–316.
Canty, A. and B. Ripley. 2010. boot: Bootstrap R (S-Plus) Functions.
R package version 1.2-42.
Cohen, J.A. 1978. Cuon alpinus. Mamm. Spec. 100: 1–3.
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
K.E. Jenks et al.: Thailand dholes
DeMatteo, K.E. and B.A. Loiselle. 2008. New data on the status
and distribution of the bush dog (Speothos venaticus): evaluating its quality of protection and directing research efforts. Biol.
Conserv. 141: 2494–2505.
Durbin, L.S., A. Venkataraman, S. Hedges and W. Duckworth. 2004.
Dhole. In: (C. Sillero-Zubiri, M. Hoffman and D.W. Macdonald,
eds.) Canids: foxes, wolves, jackals and dogs – 2004 status survey and conservation action plan. IUCN/SSC Canid Specialist
Group, Gland, Switzerland, pp. 210–219.
Elith, J., C.H. Graham, R.P. Anderson, M. Dudik, S. Ferrier, A.
Guisan, R.J. Hijmans, F. Huettmann, J.R. Leathwick, A. Lehmann,
J. Li, L.G. Lohmann, B.A. Loiselle, G. Manion, C. Moritz, M.
Nakamura, Y. Nakazawa, J.M. Overton, A.T. Peterson, S.J.
Phillips, K. Richardson, R. Scachetti-Pereira, R.E. Schapire, J.
Soberon, S. Williams, M.S. Wisz and N.E. Zimmermann. 2006.
Novel methods improve prediction of species’ distributions from
occurrence data. Ecography 29: 129–151.
Elith, J., S.J. Phillips, T. Hastie, M. Dudik, Y. En Chee and C.J. Yates.
2011. A statistical explanation of MaxEnt for ecologists. Divers.
Distrib. 17: 43–57.
Grassman, L I., M.E. Tewes, N.J. Silvy and K. Kreetiyutanont.
2005. Spatial ecology and diet of the dhole Cuon alpinus
(Canidae, Carnivora) in north central Thailand. Mammalia 69:
11–20.
International Union for Conservation of Nature and Natural
Resources. 2008. IUCN Red List of Threatened Species. http://
www.IUCNredlist.org.
Inverarity, J.D. 1901. A wild dog’s earth. J. Bombay Nat. Hist. Soc.
7: 529–530.
Jenks, K.E., P. Chanteap, K. Damrongchainarong, P. Cutter, P.
Cutter, T. Redford, A.J. Lynam, J. Howard and P. Leimgruber.
2011. Using relative abundance indices from camera-trapping
to test wildlife conservation hypotheses – an example from
Khao Yai National Park, Thailand. Trop. Conserv. Sci. 4:
113–131.
Jennings, A.P. and G. Veron. 2011. Predicted distributions and ecological niches of 8 civet and mongoose species in Southeast Asia.
J. Mammal. 92: 316–327.
Johnsingh, A.J.T. 1981. Ecology and behaviour of the dhole or
Indian wild dog – Cuon alpinus Pallas 1811, with special reference to predator-prey relations at Bandipur. World Wildlife Fund
– Indian and International Appeals, Bangalore, India.
Johnsingh, A.J.T. 1983. Large mammalian prey-predators in
Bandipur. J. Bombay Nat. Hist. Soc. 80: 1–57.
Johnsingh, A.J.T. 1985. Distribution and status of dhole Cuon alpinus Pallas, 1811 in South Asia. Mammalia 49: 203–208.
Johnsingh, A.J.T. 1992. Prey selection in three large sympatric carnivores in Bandipur. Mammalia 56: 517–526.
Johnsingh, A.J.T. and K. Yoganand. 1999. Range overlap in
dhole Cuon alpinus Pallas and wolf Canis lupus Linn.
(Family: Canidae) in India. J. Bombay Nat. Hist. Soc. 97:
418–419.
Kanchanasaka, B. 2005. Status and distribution of large mammals
in Thailand. Thailand Department of National Parks, Wildlife
and Plant Conservation. Bangkok, Thailand. ISBN: 974-992782-9.
Karanth, K.U. and M.E. Sunquist. 2000. Behavioural correlates of
predation by tiger (Panthera tigris), leopard (Panthera pardus)
and dhole (Cuon alpinus) in Nagarahole, India. J. Zool. 250:
255–265.
Karanth, K.K., J.D. Nichols, J.E. Hines, K.U. Karanth and N.L.
Christensen. 2009. Patterns and determinants of mammal species
occurrence in India. J. Appl. Ecol. 46: 1189–1200.
9
Kawanishi, K. and M.E. Sunquist. 2008. Food habits and activity patterns of the Asiatic golden cat (Catopuma temminckii) and dhole
(Cuon alpinus) in a primary rainforest of Peninsular Malaysia.
Mamm. Study 33: 173–177.
Liu, C., P.M. Berry, T.P. Dawson and R.G. Pearson. 2005. Selecting
thresholds of occurrence in prediction of species distributions.
Ecography 28: 385–393.
Lynam, A.J., K. Kreetiyutanont and R. Mather. 2001. Conservation
status and distribution of the Indochinese tiger (Panthera
tigris Corbetti) and other large mammals in a forest complex in northeastern Thailand. Nat. Hist. Bull. Siam Soc. 49:
61–79.
Lynam, A.J., C. Kanwatanakid and C. Suckaseam. 2003. Ecological
monitoring of wildlife at Khao Yai National Park, Thailand. Final
Report submitted to Department of National Parks, Wildlife and
Plants and Khao Yai Conservation Project. 64 pp.
Meyer, D., A. Zeileis and K. Hornik. 2010. vcd: Visualizing categorical data. R package version 1.2-9.
Monterroso, P., J.C. Brito, P. Ferreras and P.C. Alves. 2009. Spatial
ecology of the European wildcat in a Mediterranean ecosystem:
dealing with small radio-tracking datasets in species conservation. J. Zool. 279: 27–35.
Morris, R.C. 1925. Wild dogs and jungle tragedies. J. Bombay Nat.
Hist. Soc. 30: 691–693.
Ngoprasert, D., A.J. Lynam, R. Sukmasuang, N. Tantipisanuh,
W. Chutipong, R. Steinmetz, K.E. Jenks, G.A. Gale, L.I.
Grassman, Jr., S. Kitamura, J. Howard, P. Cutter, P. Cutter, D.H.
Reed. In press. Occurrence of three felids across a network of
protected areas in Thailand: prey, intraguild, and habitat associations. Biotropica.
Papes, M. and P. Gaubert. 2007. Modelling ecological niches from
low numbers of occurrences: assessment of the conservation status of poorly known viverrids (Mammalia, Carnivora) across two
continents. Divers. Distrib. 13: 890–902.
Pattanavibool, A. and P. Dearden. 2002. Fragmentation and wildlife
in montane evergreen forests, northern Thailand. Biol. Conserv.
107: 155-164.
Peploe, W.C.G. 1947. Field notes on the mammals of South
Tinnevelly, South India. J. Bombay Nat. Hist. Soc. 46:
629–644.
Phillips, S.J. and M. Dudik. 2008. Modeling of species distributions
with Maxent: new extensions and a comprehensive evaluation.
Ecography 31: 161–175.
Phillips, S.J., R.P. Anderson and R.E. Schapire. 2006. Maximum
entropy modeling of species geographic distributions. Ecol.
Modell. 190: 231–259.
Phythian-Adams, E.G. 1949. Jungle memories part IV – wild dogs
and wolves, etc. J. Bombay Nat. Hist. Soc. 48: 645–655.
Prater, S.H. 1965. The book of Indian animals. Bombay Natural
History Society, Madras, pp. 130–131.
R Development Core Team. 2010. R: a language and environment
for statistical computing. R Foundation for Statistical Computer,
Vienna, Austria. ISBN 3-900051-07-0, http://www.R-project.org.
Accessed on April 2011.
Rice, C.G. 1986. Observations on predators and prey at Eravikulam
National Park, Kerala. J. Bombay Nat. Hist. Soc. 83: 283–305.
Salangsingha, N. and P. DoungKae. 2009. Prey of dhole (Cuon alpinus) at Phu Khieo Wildlife Sanctuary. J. Forest Biol. (Thailand)
34: 54–64.
Sing, T., O. Sander, N. Beerenwinkel and T. Lengauer. 2009. ROCR:
Visualizing the performance of scoring classifiers. R package
version 1.0-4. http://CRAN.R-project.org/package = ROCR.
Accessed on April 2011.
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM
10 K.E. Jenks et al.: Thailand dholes
Singh, R., P.K. Joshi, M. Kumar, P.P. Dash and B.D. Joshi. 2009.
Development of tiger habitat suitability model using geospatial
tools – a case study in Achankmar Wildlife Sanctuary (AMWLS),
Chhattisgarh India. Environ. Monit. Assess. 155: 555–567.
Stewart, P. 1993. Mapping the dhole. IUCN/SSC Canid Specialist
Group. Canid News 1: 18–21.
Swets, J.A. 1988. Measuring the accuracy of diagnostic systems.
Science 240: 1285–1293.
Venkataraman, A.B. 1995. Do dholes (Cuon alpinus) live in packs
in response to competition with or predation by large cats? Curr.
Sci. 69: 934–936.
Wilting, A., A. Cord, A.J. Hearn, D. Hesse, A. Mohamed, C. Traeholdt,
S.M. Cheyne, S. Sunarto, M.A. Jayasilan, J. Ross, A.C. Shapiro,
A. Sebastian, S. Dech, C. Breitenmoser, J. Sanderson, J.W.
Duckworth and H. Hofer. 2010. Modelling the species distribution of flat-headed cats (Prionailurus planiceps), an endangered
South-East Asian small felid. PLoS One 5(3): e9612.
Wood, H.S. 1929. Observations on the wild dog. J. Darjeeling Nat.
Hist. Soc. 4: 7–15.
Woodroffe, R. and J.R. Ginsberg. 2005. King of the beasts? Evidence
for guild redundancy among large mammalian carnivores. In:
(J.C. Ray, K.H. Redford, R.S. Steneck and J. Berger, eds.) Large
carnivores and the conservation of biodiversity. Island Press,
Washington, DC. pp. 527.
Received June 21, 2011; accepted February 6, 2012
Brought to you by | University of Massachusetts - Amherst W.E.B. Du Bois Library (University of Massachusetts - Amherst W.E.B. Du Bois Library)
Authenticated | 172.16.1.226
Download Date | 4/24/12 4:34 PM